In state and local government, forecasting in tax revenue is the first step in the budgetary decision making process. It is supposed to set the budgetary limits within which public spending should remain in order to reach a fiscal balance. Although this step appears to be a critical one, most governments face difficulties in forecasting with any degree of preciseness. This may lead to a different fiscal situation than the expected one. Furthermore a trend towards under- rather than overestimation can generally be observed. As an explanation of this tendency, previous studies assert that underestimation of expected tax revenues may be strategic behaviour on the part of the forecasters to curb the financial appetite of the other actors in the budgetary process.
The aim of this research is to analyse the relationship between tax revenue forecasting and governments' fiscal position. More specifically the first step consists in identifying of the forecasting error and determining whether these errors have a systematic component (1st hypothesis). There will follow an analysis of whether a systematic under-estimation of tax revenues does in fact generate a positive effect on the fiscal position (2nd hypothesis) as has been assumed in the literature but has never yet been empirically tested. Should this positive effect be verified, it would mean that a cautious strategy of under-estimation is generally successful.
The test of the first hypothesis consists in the assessment of the quality of tax revenue forecasting through different tax revenue aggregates. These aggregates are then regressed through a wide range of financial, fiscal, economic and political variables using a parametric model. In order to assess the robustness of the results, the first hypothesis is then estimated using a non parametric model.
The effect of forecasting error in tax revenue on the fiscal situation is then estimated using a simultaneous equation model of state revenue and spending. According to the theory, a negative correlation between the underestimation error variable and the spending variable is expected. The second hypothesis is also tested using a single equation model of fiscal balance. However we expect the first estimated model to give better results.